FOR NICOHWILLIAM ONLY

The Timing of Railway Construction on the Canadian Prairies Author(s): Frank D. Lewis and David R. Robinson Source: The Canadian Journal of Economics / Revue canadienne d'Economique, Vol. 17, No. 2 (May, 1984), pp. 340-352 Published by: Wiley on behalf of the Canadian Economics Association Stable URL: http://www.jstor.org/stable/134961 Accessed: 01-03-2017 19:21 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted

digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about

JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

http://about.jstor.org/terms

Canadian Economics Association, Wiley are collaborating with JSTOR to digitize, preserve and extend access to The Canadian Journal of Economics / Revue canadienne d'Economique

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

The timing of railway construction

on the Canadian Prairies

FRANK D. LEWISand DAVID R. ROBINSON/

Queen's University

Abstract. It is generally agreed that railways were a prerequisite to the Canadian 'wheat boom.' In this paper we analyse the timing of railway construction in south-eastern Saskatchewan during the wheat boom period. A monopoly model is derived in which a railway makes its construction and pricing decisions on the basis of a rational settler's response function. Between 1898 and 1906, when the CPR had a monopoly, the model closely predicts freight rates and the timing of branch line completions. The increase in branch line construction after 1898 is explained partly by the decline in the railway's operating cost and partly by the Crow's Nest Pass Agreement. This agreement raised the value of CPR land by setting long-term ceilings on freight rates.

Le moment approprie pour la construction du chemin defer dans les prairies canadiennes. On est generalement d'accord pour dire que les chemins de fer ont ete une condition necessaire pour le boom du ble. Dans ce memoire, les auteurs analysent les decisions quant au moment de la construction des chemins de fer dans le sud-est de la Saskatchewan au cours de la periode du boom du ble. Ils derivent un modele de monopole dans lequel la compagnie de chemin de fer prend ses decisions quant a la construction et quant 'a la tarification sur la base d'une fonction de reaction des colons rationnels. Entre 1898 et 1906, la periode de monopole du CPR, le modele permet de predire 'a quel moment on devrait completer la construction de certaines lignes secondaires et quelle devrait etre la tarification. L'acceleration de la construction des lignes secondaires apres 1898 est attribuable en partie a la chute dans les couits d'operation du CPR et en partie a I'Accord des tarifs du Nid de Corbeau. Cet accord elevait la valeur des terres du CPR en imposant un plafond 'a long terme sur les tarifs.

INTRODUCTION

Economic historians have long been puzzled by the timing of Canadian prairie

settlement. The Canadian Pacific Railway's transcontinental line was completed in 1885, eleven years before the beginning of the 'wheat boom'; and it was not until after

1900 that there was large-scale settlement west of the Manitoba-Saskatchewan

border. In a series of recent articles economic historians have tried to explain this lag,

For their helpful comments we thank Alan Green, Knick Harley, Marvin McInnis, and Gillian Wogin. An earlier version of this paper was presented to the Twelfth Conference on Quantitative Methods in Canadian Economic History at Edmonton, 1982.

Canadian Journal of Economics / Revue canadienne d'Economique XVII, No. 2 May / mai 1984. Printed in Canada / Imprime au Canada

0008-4085 / 84 / 340-352 $01.50 ?) 1984 Canadian Economics Association

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

The timing of railway construction / 341

and although no consensus has emerged on the causes of settlement, progress has

been made. Marr and Percy (1978) point to large expenditures by the Canadian

government to promote immigration and to the increase in railroad mileage. Norrie

(1980) finds rapid diffusion of dry-farming techniques; and Dick (1982) estimates a

gradual decline in the cost per acre of growing wheat. In another recent paper, Lewis

(1981) develops a simulation model of prairie settlement. He finds that settlement

advanced with the feasible region of wheat cultivation, the area on which wheat

farming was profitable.' Increasing mean wheat yields accounts in part for the extension of the feasible region, but it is the construction of railway branch lines that has by far the larger impact.

In all these papers, settlement is analysed on the basis of an exogenously given rail

network. Missing is an account of the railway's decision to construct lines. To fill this

gap we develop a model that recognizes explicitly the interdependence of the

railway's and settlers' decisions and treats the timing of railroad construction as

endogenous. At the heart of our model is the hypothesis that railway companies know

the response function of settlers. The firms, therefore, base their construction

decision not on observed settlement but rather on their estimate of future settlement

conditional on a rail line being completed.2 We apply our model to south-eastern

Saskatchewan during a period when there was no interrailway competition - the CPR had a monopoly - and implement the notion of a settler's response function by using a

simplified vesion of the Lewis (1981) model, which he tested on the same region.

We begin by outlining briefly the pattern of early railway construction in south-eastern Saskatchewan. In addition to the transcontinental line, the CPR

completed two lines prior to 1900: a branch from Moose Jaw to North Portal on the

u.s. border and a section that connected Esteven and Souris, Manitoba (see figure 1).

Neither of these lines, however, can be explained on the basis of locally generated

traffic. They were through lines designed to provide a connection to the u.s. market in

Chicago. Two colonization railroads that passed through the region also were

completed before 1900: the Qu'Appelle Long Lake and Saskatchewan Railway (QLL)

that extended north from Regina to Prince Albert, and the Manitoba and North

Western Railway (MNW) that lay between Portage la Prairie and Yorkton. Like other

colonization railways they were subsidized with large land grants but were not

successful at attracting many settlers. In fact, the poor performance of the

colonization railways led Hedges (1934) to conclude: 'more interested in securing land than in operating railways, their every move was calculated to obtain the

maximum amount of the best land, with a minimum expenditure of money and effort' (69). In 1890 the CPR began operating the QLL, and when in 1900 it leased the MNW, it

secured a monopoly on rail transportation in Saskatchewan. Its position was maintained until 1905, when the Canadian Northern purchased the QLL. Then in 1907

1 Lewis applies his model to one area in Saskatchewan: Crop District 1 (1908-14 definition), a district of about 16,800 square miles in the south-eastern corner of the province.

2 In his book on the history of the CPR, Innis (1971) pointed out the interdependence of the decisions made by the railway and the settlers both with regard to the timing of branch line construction and the setting of freight rates. See, in particular, the discussions on pages 129-30 and 172-9.

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

342 / Frank D. Lewis and David R. Robinson

_ ~~~YORKTON ?SAE s ' /5 l ~~35 M ILES TO 3 I NCH

STRASSBURG /

- - -QPATRICK

FREGINA,

N ARCOLA,- z~ U I - \\"\

ESTE VA -PORTAL

LINES COMPLETED BEFORE 1898

- - - LINES COMPLETED FROM 1898 TO 1906

LINES COMPLETED FROM 1907 TO 1911

FIGURE 1 Rail lines: Crop District 1, Saskatchewan

the CPR'S monopoly was further weakened as both the Canadian Northern and Grand Trunk Pacific completed main lines through the region. In fact, after 1907 the railway industry in Saskatchewan is best characterized as an oligopoly, and the monopoly model that we develop in the next section no longer applies.

A MONOPOLY MODEL OF RAILWAY CONSTRUCTION

A railway is assumed to secure a monopoly in a region of the frontier by constructing the only rail line through that region. Although completion of the main line may be premature based on current traffic and any subsidies that the railway receives, the railway anticipates that profits from future branch lines will be at least enough to offset current losses.3 Once the main line is completed, the firm's objective is to delay

3 Harley (1982, 797-805) has an excellent discussion of the economic factors affecting the timing of

main line and branch line construction. Indeed, one contribution of our paper is to formalize Harley's treatment.

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

The timing of railway construction / 343

the construction of branch lines until the net present value of each of these lines is

maximized. Income from a branch line is assumed to come from two sources: net

revenue from the rail traffic generated and rent on any railway land that is served by

the line.4 Formally, the firm maximizes the net present value of each branlch line with

respect to completion time and the freight rates that it charges once the line is opened:

max rT {{[Pr(t) - Cr(t)]F[t,Pr(t)] + PR[t,Pr(t)] jdt

Pr(t), T

- [(r + 8)lr]pke-, (1)

where Pr = rail freight rate, Cr = railway's operating cost, F = rail traffic generated by the branch line, p = proportion of land owned by the firm, R = effect of the line on

land rents, Pk = cost of constructing the line, r = discount rate, 8 = depreciation rate

of the line (this is assumed to be offset by maintenance and improvements to the line),

and T = optimal construction time (from the viewpoint of the firm). The first-order

condition for the optimal construction time is

[Pr(T) - Cr(T)]F[T, Pr(T)] + pR[T, Pr(T)] = (r + 8)Pk. (2)

The intuition for equation (2) is straightforward. It requires that the railway construct

each branch line when the net revenue from rail traffic plus the increased rent on railway land equals the interest and depreciation on the road.5 The optimal freight rate

in each period is the solution to the following equation, which is derived by

differentiating equation (1) with respect to Pr(t)

F[t, Pr(t)] + p{dR[t, Pr(t)J/dPr(t)} + [Pr(t) - Cr(t)]{dF[t, Pr(t)]ldPr(t)} = 0.6 (3)

We derive both the rail traffic function, F, and the rent function, R, from the

feasible region of wheat cultivation.7 In other words, we assume that the railway

bases its freight rate and timing decisions on the assumption that settlers will occupy

all land where wheat. farming is profitable. The margin of wheat cultivation establishes the boundary of this region. The boundary lies at that distance from a rail

line where the exogenously given rnarket price of wheat equals the cost of producing

the wheat and transporting it to market by wagon and rail:

P(t) = cw(t) + Pr(t) + Ph(t)-D[t, Pr(t)], (4)

where p, = price of wheat, c, = cost of producing wheat (exclusive of land rent), Ph

4 Although railways seldom rented their land, land rents are assumed to be capitalized into the value of the land.

5 Since the annual cost of the line is constant, the second-order condition is that net income is rising at construction time.

6 In other words, the railway raises its freight rate until the increase in revenue due to the higher rate is just offset by a decline in income due to reduced rail traffic and land rent.

7 Our approach is based on the Lewis (1981) model with two simplifying assumptions. First, we assume that wheat yield is the same across all wheat land. Lewis assumed a normal distribution of yields. Second, we allow wheat to be shipped from any point along the rail line. Lewis required wheat to be shipped from loading platforms that were spaced about eight miles apart. Neither assumption seriously biases the results.

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

344 / Frank D. Lewis and David R. Robinson

= wagon haulage rate per mile, Pr = rail freight rate, and D is the distance from the line in miles where land rent is zero (all prices and costs are per bushel). Rearranging equation (4) gives

D[t, pr(t)] = [Pw(t) - Cw(t) - Pr(t)]Ph(t)* (5)

If there is no overlap with other rail lines, each mile of track serves an area of 2D square miles.

Wheat production in the feasible region provides the basis for computing the increase in rail traffic:

F[t, pr(t)] = 2D[t, Pr(t)]Y(t)f, (6)

where y = wheat yield per acre, k is a factor that converts square miles into wheat acreage, and f is a factor that converts wheat production into rail traffic (in wheat equivalents). The increase in land rent is also based on the output of wheat:

R[t, Pr(t)] = 2D[t, Pr(t)]ky(t) [ 2[pP,(t) - CW(t) - Pr(t)]1M, (7)

where m is a factor that converts rent on wheat land to rent on all land. Note that since

rent per bushel of wheat varies linearly from 0 (at the margin of cultivation) to pw - Cw - Pr (at the rail line), average rent is 2(pw - CW - Pr). Finally, substituting equations (5), (6), and (7) into equation (3), we derive the optimal freight rate to be

Pr*(t) = [fl(2f - pm)] {[1 - p(f/m)] [pw(t) - cj(t)] + cr(t)}. (8)

The derivation of the optimal freight rate can also be illustrated diagrammatically. In figure 2, A represents the demand curve for rail services and B the railway's average revenue curve, which lies above A if the railway owns land in the area served by the line. The firm maximizes profits by setting marginal cost equal to marginal revenue, based on the B curve. Note that if the firm owns all the land (and adjustment factors m and f are equal), the slope of A is twice the slope of B and the firm acts as a perfect price discriminator, setting price equal to marginal cost. If on the other hand

the firm owns no land, it charges the monopoly price, 2(pw - Cw - Cr).8

AN APPLICATION OF THE MODEL TO CROP DISTRICT 1,

SASKATCHEWAN

We now apply the model to the construction of branch lines by the CPR in Crop District 1, Saskatchewan from 1898 to 1905. During the period the CPR laid three sections of track (see table 1). A 64-mile branch from Reston to Arcola was completed in 1901; a 140-mile section of the Pheasant Hills branch, from Kirkella to Patrick, was

8 Engerman (1972) discussed some theoretical issues concerning the subsidies provided to u.s. rail- roads. He pointed out that giving a subsidy in the form of a land grant not only could lead to increased railroad construction but also would result in lower freight rates. In a recent PH D dissertation, Gillian Wogin (1983) explores the relationship between freight rates and land settlement policy in Canada. She finds that land grants to the CPR induced the railway to lower its freight rate from the monopoly price to a more efficient level.

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

The timing of railway construction / 345

Pr

PW - S

Pr*

A \B

Cr

I \\MRB

0

F* F, F2 Rail traffic

F, = [2(pw - CW)kyf/Ph-

F2 = -4(pw cw)kyf2]1[ph(2f- pm)].

F* = [2(pw- w- Cr)kyf2]I[Ph(2f - pm)].

FIGURE 2 Derivation of the optimal freight rate

completed in 1904; and in 1905 the Reston-Arcola line was extended 113 miles to

Regina (see figure 1). We analyse the timing decisions by comparing the annual cost

of each section of track with the income that each section was expected to generate during its first year of operation based on the actual freight rate (see equation (2); the effect on the calculations of using optimal freight rates is discussed below). The capital cost of these lines was among the lowest of all track laid by the CPR on the Prairies - about $12,000 per mile.9 Multiplying by the sum of the discount rate and the depreciation rate of the road, put at 6 and 4 per cent, respectively, gives an annual cost of about $1,200 dollars per mile of track. 0 It is this amount that we compare with net income from rail traffic and land rents. To derive net revenue from rail traffic, we first compute the effect of each branch line on the feasible region of cultivation. Net revenue is simply the implied increase in the volume of traffic times the difference

9 The extension of the Pheasant Hills branch from Strassburg to Hardisty, for example, cost about

$20,000 per mile, and the 270-mile Moose Jaw Branch, from Moose Jaw north-westerly to Macklin, cost $18,000 per mile.

10 Depreciation is assumed to have been offset by maintenance of track and structures. For all CPR lines the ratio of the cost of maintenance to the capital value of road and structures was about 4 per cent. The capital value is derived by cumulating all the (deflated) expenditures of the CPR on construction and improvements (CPR, Annual Reports).

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

346 / Frank D. Lewis and David R. Robinson

TABLE 1

The cost of constructing branch lines in Crop District 1, Saskatchewan

Year Distance Cost per milea Line completed (miles) (1890-99$)

1. Reston to Arcola (Pipestone Extension) 1901 64 12,400

2. Kirkella to Patrick (Pheasant Hills Branch) 1904 140 12,000 3. Arcola to Regina (Pipestone Extension) 1905 113 12,400 4. Reston to Kaiserb (Wolseley Branch) 1908 122 12,200

SOURCES: CPR, Annual Reports, 1898 to 1908, appendices; Urquhart and Buckley (1965), 292. aAll construction expenditures are deflated to 1890-99 dollars and compounded at 6 per cent to the year in which the line was completed. 'This line was completed after the CPR lost its monopoly in the district. It is included so that we can compare the monopoly with the oligopoly case.

between the CPR'S freight rate and its operating cost. Finally, net income is obtained

by adding to this amount the increase in land rent.

We illustrate the calculation for the Reston-Arcola section, which was completed in 1901 (see tables 2 and 3). In that year the extensive margin of cultivation (D) was

11.8 miles from the rail line and the feasible region per mile of track was 19.4 square

miles.11 Average wheat output per square mile (ky) was 1,690 bushels of which 78 per cent was marketed. 12 But wheat generated additional rail traffic in the form of other agricultural output, imported freight, and passengers. This traffic provided 90

per cent of the revenue obtained from carrying wheat. We assume, therefore, that the

CPR transported the equivalent of 1.48 bushels of wheat for each bushel of wheat

produced. 13 The revenue derived from this traffic is based on the CPR'S freight rate to Fort William, which in 1901 was the Crow rate of 10.1? per bushel. The CPR'S

11 From table 2, D = (Pw - C, - Pr)'Ph = (64.3 -47.5 - 10.1)/0.567. The region served by the Reston-Arcola line was less than 2D square miles, because part of the track passed just south of the

Moose Mountain Forest Reserve, an area unsuited for growing wheat. 12 Lewis (1981, 529-30) estimated that 17.7 per cent of Crop District 1 was suitable for growing wheat

and that mean wheat yield in 1901 was 14.94 bushels (case 3 of his model). (Note that there are 640

acres in a square mile.) The proportion of wheat output transported by rail is based on a comparison

of the quantity of wheat inspected with the quantity produced over the period 1908 to 1911 (Saskat- chewan, Department of Agriculture, Annual Reports).

13 A breakdown of rail traffic is available only by railway. To obtain an estimate of the traffic trans- ported on branch lines in south-eastern Saskatchewan we selected a small railway that served a comparable area. This was the Brandon, Saskatchewan and Hudson Bay Railway, a sixty-nine-mile line about sixty-five miles east of Crop District 1 that connected Brandon, Manitoba to Bannerman, near the U.S. border. Railway Statistics reports the tonnage of grain and other types of freight

transported along this line. It is assumed that 84.2 per cent of grain tonnage consisted of wheat. This figure is based on output in Crop District 1 and a comparison of output and grain inspections for all of

Saskatchewan - in converting bushels to tons the following weights are applied: wheat: 60 lbs; oats: 35 lbs; barley: 48 lbs; flax: 56 lbs. We estimate that the ratio of wheat and flour tonnage to the

tonnage of all freight was 0.586. Finally, assuming that 10 per cent of railway revenue was generated by passenger traffic, we obtain a ratio of wheat traffic to all rail traffic of 0.527. Since 78.1 per cent of wheat output was marketed, it follows that the equivalent of 1.48 bushels of wheat was transported

for every bushel of wheat produced. (These estimates are based on 1908- 11 averages; see Canada, Department of Railways and Canals, 1908-11; Saskatchewan, Department of Agriculture, 1908-

11.)

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

The timing of railway construction / 347

TABLE 2

Data on the prices and costs of wheat production in Crop District 1, Saskatchewan (1890-99 cents per bushel)

1898 1901 1904 1905 1908

Price of wheata 63.7 64.3 64.9 65.1 65.7 Cost of producing wheatb 48.8 47.5 49.5 49.7 46.6 Rail freight rate to Fort Williamc 10.1 8.6 9.0 7.9 Mean wheat yieldd(bushels per acre) 14.95 14.94 15.03 15.24 16.99 Wagon haulage rate per milee 0.567 0.567 0.567 0.567 0.567

aThis is the (quality-adjusted) trend price of wheat at Fort William between 1898 and 1911 less 3? per bushel for commissions and handling (Lewis, 1981, 525-6). bLewis (1981) separated post-harvest costs, which were assumed to depend on yield, from the other costs of producing wheat. Here we combine the two types of costs, but otherwise the cost estimates are the same (524-5). cFreight rates are given for each line in the year they were completed: 1901: Reston-Arcola; 1904: Kirkella-Patrick; 1905: Arcola-Regina; 1908: Reston-Kaiser. The freight rates are based on the average distance of each line from Fort William (McDougall, 1966). dBased on case 3 of Lewis (1981, 530) eSee Lewis (1981, 526-7).

operating cost of moving this wheat to Fort William was 7.0? per bushel, implying that net earnings per bushel were 3. 1 ?. Combining these values, it follows that on the

Reston-Arcola section in 1901 net revenue from rail traffic was $1,500 per mile of

track (0.031 x 1.48 x 1,690 x 19.4).

Net revenue from rail traffic understates the full impact of the branch line on CPR

income, because the firm owned a large share of the land in Crop District 1. In nearly

every township the CPR had been granted the odd-numbered sections, less two sections that were designated as School Lands. Since there are thirty-six sections in a

township, it follows that 44 per cent of the land in the district was owned by the CPR. In 1901 the Reston-Arcola line increased the rent on wheat land by $1,100 per mile of track. 14 This estimate, however, understates the impact of the line on all rents, because agricultural land also was used to raise livestock and to grow oats, barley, and other crops. Adjusting for these rents and for the fact that the CPR owned less than

half the land, we estimate that per mile of track the Reston-Arcola line increased the rent on CPR land by $850.15 Combining land rent with net revenue from rail traffic and comparing the sum with the annual cost of the line, we conclude that, based on

our model, the CPR should have completed the Reston-Arcola section prior to 1901. In

1901, the road imposed an annual cost of $1,240 per mile but provided income of

$2,350: $1,500 from rail traffic and $850 as the increase in the rent on CPR land. The timing of the Kirkella-Patrick and Arcola-Regina sections conforms much

14 R = rent per bushel x wheat output per square mile x feasible region = -(p,-c-Pr) x 1,690 x 19.4 = 1,100.

15 The ratio of rent on all land to rent on wheat land, m, is estimated to be 1.76. This is the Crop District 1 average for 1908-1 1 and is based on the value of all field crops and the value of output generated by some livestock (milk cows, other cattle, sheep). Saskatchewan, Department of Agriculture, 1908- 11.

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

348 / Frank D. Lewis and David R. Robinson

TABLE 3

Annual costs and returns of Railway branch lines: Crop District 1, Saskatchewan (1890-99 prices)

Reston- Kirkella- Arcola- Reston-

Arcola Patrick Regina Kaiser (1901) (1904) (1905) (1908)

1. Annual costa (dollars per mile) 1,240 1,200 1,240 1,220

2. Operating costb (cents per bushel - 7.0 7.2 7.4 6.9 cents per ton-mile in parentheses) (0.367) (0.342) (0.344) (0.355)

3. Freight rate (cents per bushel - 10.1 8.6 9.0 7.9 optimal rate in parentheses) (10.2) (9.8) (10.0)

4. Area servedc (sq. miles per track mile) 19.4 16.5 17.1 6.2

5. Income from rail trafficd (dollars per mile) 1,500 580 700 180

6. Income from land rente (dollars per mile) 850 740 730 35f

7. Annual income/annual cost 1.90 1.10 1.15 0.18

aO.I x capital cost (see table 2 and fn. 10). bOperating cost per ton-mile is based on the sum of the reported costs of conducting transportation, motive power (when given), and the maintenance of rolling stock. Also included are depreciation and interest on rolling stock, put at 5 per cent and 6 per cent respectively (a low depreciation rate was applied because of the high expenditures on maintenance). The value of rolling stock was derived by cumulating all CPR expenditures from 1885, and applying a 5 per cent depreciation rate. The share of operating costs attributed to freight is set equal to the share of income from freight. Operating cost per bushel of wheat is based on the distance to Fort William and a weight of 60 lb per bushel (CPR, Annual Reports). cBased on equation (5) and table 2. The results are adjusted downward to take account of areas that are not suited for wheat-growing and overlap with other rail lines. d[pr(t) - cW(t)] area served y(t) kf, where k is 113 wheat acres per square mile and f, the ratio of rail traffic to wheat output, is 1.48 (see fn. 13).

e2[p(t) - C.(t) - Pr(t)] area served y(t) . kmp, where m, the ratio of rent on all land to rent on wheat land, is 1.76; and p, the share of land owned by the CPR, iS 0.44 (see fn. 15). fSince nearly all land served by the Reston-Kaiser section would have been in the feasible region without the line, the effect on land rent is small.

more closely to the predictions of the model. These lines, completed in 1904 and 1905, respectively, generated less income per mile of track than the Reston-Arcola section because they overlapped with other lines and hence served smaller areas. In 1904, its first year of operation, the Kirkella-Patrick section imposed an annual cost of $1,200 but generated net revenue from rail traffic of only $580 (see table 3). Nevertheless, the line was completed at about the optimal time because the shortfall was more than offset by an increase in the rent on CPR land. The timing of the Arcola-

Regina section was also close to optimal. In 1905, when that line was completed, the annual cost was $1,240 and net income, including land rent, was $1,430. An important implication of these results is that the land grants to the CPR accelerated branch line construction, even though they were mainly intended to subsidize the transcontinental line. Had the CPR owned no land, completion of both the Kirkella-Patrick and Arcola-Regina sections would have been delayed. In fact, when these lines were first opened, the CPR derived more income as capitalized land rent than as net revenue from freight and passengers.

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

The timing of railway construction / 349

TABLE 4

Hypothetical costs and returns of railway branch lines (1890-99 prices)'

Reston-Arcola Kirkella-Patrick Arcola-Regina

A B A B A B

1. Operating costb 7.7 7.7 8.5 8.5 8.7 8.7 2. Freight rate 13.1 10.3 13.1 10.6 13.7 10.7 3. Area served 6.3 14.1 5.9 12.4 4.0 12.8 4. Income from rail traffic 850 920 680 650 500 640 5. Income from land rent 75 425 70 350 30 350 6. Annual income / annual cost 0.75 1.08 0.62 0.83 0.43 0.80

A = based on 1898 values. B = based on 1898 values but assuming the optimal freight rate in that year. aFor the units in which variables are expressed, see table 3. bIn 1898 operating costs were 0.403? per ton-mile (see table 3, fn. b)

On the basis of our model, we accurately predict the timing of two of the three rail lines in our study but estimate that construction of the third, the Reston-Arcola

section, was delayed. The Reston-Arcola section was completed in 1901, when net income provided by the line far exceeded its annual cost. Surprisingly, however the

lag between the optimal and actual completion time was at most two years. We

establish this by recomputing our results using data for 1898 (see table 4). In that year the Reston-Arcola section would have generated net income, including capitalized land rent, of $925, which is only three-quarters of the annual cost of the road. It was optimal, therefore, for the CPR to wait until at least 1899 before completing the line.

In addition to explaining construction time, our model can also help account for the freight rates charged by the CPR. Between 1898 and 1905 the CPR reduced its nominal freight rates twice. In 1899 the freight rate to Fort William was reduced by 3? per

cwt'6 in accordance with the Crow's Nest Pass Agreement. Then in 1903 the Manitoba Agreement lowered nominal freight rates a further 2? per cwt. We have compared actual and optimal freight rates on the three lines for the years when each line was completed. In 1901, for example, the optimal freight rate to Fort William

from the Reston-Arcola section was 10.2? per bushel (see table 3), which is almost identical to the actual rate - the Crow - of 10. 1?. On wheat shipped from the

Kirkella-Patrick and Arcola-Regina sections, optimal freight rates were 9.8? and 10.0? per bushel, respectively, in the years when those lines were completed. These rates are well below the Crow rates of 11.6? and 11.9? but are close to the rates

stipulated in the Manitoba Agreement, which by then was in effect. It should be noted

than land ownership by the CPR reduced the optimal freight rate by about 1.5? per bushel.

Our results indicate that, as a monopolist, the CPR followed a close to optimal plan with respect to both freight rates and the timing of branch line construction. Next, we

16 Freight traffic in Canada was measured on the basis of American units: 100 lbs per cwt and 2,000 lbs per ton.

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

350 / Frank D. Lewis and David R. Robinson

combine our findings with those of Lewis (1981), to help complete our explanation

for the timing of the wheat boom. In his paper Lewis showed that in south-eastern

Saskatchewan large-scale prairie settlement had to await the completion of railway

branch lines. In this paper we show that the reason branch line construction did not

take place earlier was that prior to 1899 branch lines would have generated less

income per year than the annual cost of the lines. After 1899 conditions changed and

the laying of track became the profit-maximizing course to follow.

Two factors contributed to the increased profitability of prospective branch lines:

the Crow's Nest Pass and Manitoba agreements and a decline in the operating costs of

the CPR. We have shown that the Crow and Manitoba agreement rates were close to

optimal; however, prior to 1899, when the Crow agreement came into effect, the CPR

charged much higher freight rates. An immediate question arises: why did the CPR

wait for the implementation of the Crow's Nest Pass Agreement to lower its rates? We

offer the following answer, first suggested by Gillian Wogin (1983). When settlers

purchased land from the CPR, the price that they paid was determined not by current

wheat prices and costs of production, but rather by expected future prices and costs.

Without the Crow's Nest Pass Agreement or one like it, settlers expected that once

they purchased the land, the railway would raise its freight rate in order to extract

more monopoly rent. This meant that the lowering of current freight rates would have

had little effect on the capitalized value of CPR land, which led to suboptimal pricing.

It was only after the CPR was in a position of precommitment that reductions in freight

rates would significantly affect land prices and thereby increase the income from

branch line construction.

The Crow's Nest Pass Agreement may have been a necessary condition for branch

line construction, but it was not a sufficient one. Even if the CPR had charged the

profit-maximizing freight rate in 1898, it would not have been optimal to construct

either the Kirkella-Patrick or the Arcola-Regina line (see table 4). In fact, even the Reston-Arcola section would have generated just slightly more income than its annual

cost. The problem was the CPR'S high operating cost. Recently Alan Green (1983) has

compiled evidence showing rapid technological change in the railway sector during

the 1890s and early 1900s. Data on the CPR'S operating costs are consistent with his

results. From the 1890s to the period 1900- 10 operating costs fell from 0.4? to about

0.35? per ton-mile. 17 This reduction may seem small, but it had a big impact on the margin between freight rates and operating costs. 8 Green has not identified the sources of railway productivity change in Canada, but Fishlow (1966, 626-31), who

also estimated large increases in total factor productivity for u.s. railroads during this period, points to the increased size of locomotives, their increased motive power per

unit weight, and the higher ratio of load to dead weight in freight cars. These changes

may have affected Canadian railways as well. On the Canadian Prairies, moreover,

improvements in productivity not only led to reductions in freight rates; they also

contributed to the extension of the rail network.

17 Operating costs are derived from the CPR, Annual Reports. See table 3, fn. b. 18 In 1905, for example, the freight rate on the Kirkella-Patrick branch exceeded operating costs by 1.4?

per bushel. But had operating costs been at their 1898 levels, this margin would have been reduced to 0. 1? (given the 1905 freight rate).

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

The timing of railway construction / 351

Our monopoly model, which accounts well for railway construction between 1898

and 1905, is no longer appropriate by 1907. In that year the industry became an

oligopoly, as the CPR, Canadian Northern, and Grand Trunk Pacific all competed for

traffic. The oligopoly case has been analysed by Harley (1982) for branch line construction in the u.s. mid-west. Harley argues that once a railway loses its

monopoly, or if an oligopoly agreement breaks down, a construction boom will

occur. Fearing pre-emptive railway construction by competitors, firms now build

lines whose net present value is positive, whereas in the monopoly case they wait until

the net present value of each line is maximized. The story Harley tells for the u.s.

mid-west also fits Crop District 1, Saskatchewan. After 1907 there was a construction

boom in which all three railways participated; moreover, the branch lines completed

during this boom were certainly built before their present values were maximized. In

1908, for example, the CPR line from Reston to Kaiser (see figure 1) imposed an

annual cost of $1,220 and generated income of only $215 (see table 3). Surprisingly,

however, the collapse of the CPR monopoly in Crop District 1 had little effect on

settlement, because enough branch lines had been completed prior to 1907 to bring

nearly all land into the feasible region of wheat cultivation (Lewis, 1981, 523). Track

laid after 1907 was for the most part redundant, in that it had almost no effect on

settlement.

CONCLUSION

As Lewis (1981) showed with a model based on the feasible region of wheat cultiva-

tion, settlement advanced not with the CPR'S transcontinental line but rather with the

construction of railway branch lines. It is, therefore, the timing of branch line

construction that is the focus of our paper. We first develop a monopoly model of railway construction that makes use of the Lewis approach. We than analyse three CPR

lines built in south-eastern Saskatchewan between 1901 and 1905. We find that the

CPR completed each section of track at about the time that maximized its net present

value. In fact, as late as 1898 it would not have been profit-maximizing to lay any of

the three sections. The Crow's Nest Pass Agreement, which came into effect in 1899, was important in raising the profitability of prospective branch lines. Since the

agreement set long-term ceilings on freight rates, it increased the capitalized value of CPR land by promising settlers low future, as well as low current, freight rates. Productivity change in the railway sector was even more important. It lowered

operating costs, reduced (optimal) freight rates, and extended the feasible region of

cultivation per mile of track. This led to increased rail traffic and land rent, which

made the construction of new rail lines profitable.

We now return to the first issue raised in this paper: the long period between the

completion of the CPR'S transcontinental line and the start of large-scale settlement. We think that the reason many economic historians have been puzzled by the long delay is that they have been posing the wrong question. Rather than ask 'why did settlement begin so late?' they should have been asking 'why was the transcontinental line completed so early?' On the basis of our model we can provide a partial answer. With the completion of the transcontinental, the CPR established a monopoly on the

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms

352 / Frank D. Lewis and David R. Robinson

provision of rail services throughout much of the Prairies. The value of this monopoly

was the expected net present value of the future branch lines that it built. The cost of

establishing the property right to these future returns was the present value of

expected losses due to premature construction. Of course huge government subsidies substantially reduced net construction costs and contributed to even earlier completion of the line. With the exception of the land grants, however, it is unlikely that these subsidies had much impact on the timing of settlement.

REFERENCE S

Canada, Department of Railways and Canals (1909-12) Railway Statistics, 1908 to 1911. In Parliament of the Dominion of Canada, Sessional Papers (Ottawa)

Canadian Pacific Railway Company (CPR), Annual Reports, 1885 to 1911 (Montreal) Dick, Trevor (1982) 'Mechanization and North American prairie farm costs, 1896-1930.'

Journal of Economic History 42, 199-206 Engerman, Stanley L. (1972) 'Some economic issues relating to railroad subsidies and the

evaluation of land grants.' Journal of Economic History 32, 443-63 Fishlow, Albert (1966) 'Productivity and technological change in the railroad sector, 1840-

1910.' In Conference on Research in Income and Wealth, Output, Employment and

Productivity in the United States after 1800. Studies in Income and Wealth, Vol. 30 (New York: NBER)

Green, Alan (1983) 'Productivity and technological change in the Canadian railroad indus- try.' Paper delivered at the Canadian Economics Association annual meeting, Vancouver

Harley, C. Knick (1982) 'Oligopoly Strategy and the Timing of American Railroad Construc- tion.' Journal of Economic History 42, 797-824

Hedges, James B. (1934) The Federal Railway Land Subsidy Policy of Canada (Cambridge, MA: Harvard University Press)

Innis, Harold A. (1971) A History of the Canadian Pacific Railway (Toronto: University of Toronto Press)

Lewis, Frank D. (1981) 'Farm Settlement on the Canadian Prairies, 1898 to 1911. ' Journal of Economic History 41, 517-25

Marr, W. and M. Percy (1978) 'The government and the rate of prairie settlement.' This JOURNAL 11, 757-67

McDougall, John Lome (1966) 'The relative level of Crow's Nest grain rates in 1899 and 1965.' Canadian Journal of Economics and Political Science 32, 46-54

Norrie, Kenneth H. (1980) 'Cultivation techniques as a response to risk in early Canadian prairie agriculture.' Explorations in Economic History 17, 386-99

Saskatchewan, Department of Agriculture (1907-12) Annual Reports, 1905 to 1911 (Regina) Urquhart, M.C. and K.A.H. Buckley, eds (1965) Historical Statistics of Canada (Toronto) Wogin, Gillian (1983) 'The wealth-maximizing behaviour of the Canadian pacific railway:

lands, freight rates, and the Crow's Nest Pass Agreement.' PH D dissertation, Carleton University

This content downloaded from 142.104.160.247 on Wed, 01 Mar 2017 19:21:49 UTC All use subject to http://about.jstor.org/terms